Euro 20-AI

For those in the European continent, most conversations are dominated by the football championship, Euro 2024. Whatever team each of you reading this article supports, let's discuss a potential future where your country's team or favourite team (there is no judgment here), is made up of ten robot players and one robot goalkeeper.

I am not saying this is a possibility...

but it seems that in our race to replace the human factor in most human interactions with AI, then I do not see why football would remain untouched. So let us assume it is happening, and then ask the question:

How is an AI football team formed?

As with every AI technology, we require Data, and a lot of it. In today's era of machine learning and AI, this data already exists but is being used to identify opponents' tactics, past mistakes, other teams' formations, and so on. The first step is already available.

Next would be to use that data for predictability, and again, that technology exists. Using Machine Learning and AI, we can predict a person's next steps based on the opponent's reaction. After all, wasn't the algorithm that beat the best chess player the one that started the AI race?

Great! Then we need robots to act out these predictabilities and again, the technology is here already! There are pet robots and machines in factories that were programmed to complete certain tasks otherwise done by humans.

In other words, we already have everything we need to build an AI team using that team's historical data. But this is where it becomes a bit complicated. If all teams built their AI teams in this manner to replace their players, the AI players would only know what to do based on past data. Without adding new data points or 'out-of-the-box' information, they will continue to re-iterate and perfect their techniques until their output is optimal based on historical data.

In order to add information, clubs start discussing trading of proprietary data because, despite the current trend that 'online data is free', which is wrong, no club will ever agree to give their data for free. I think that is something no one will question. Then the option becomes to either buy proprietary data from other teams to learn their historical data or run simulations to try and get new inputs.

The only problem with running simulations is that they will be run in a controlled environment whereby the goalkeeper is not actively trying to keep the opponent from scoring. The defender is not intuitively trying to stop the attacker, although let's face it, some don't, even now. Simulations are great for practice, but there is a massive difference between simulated data and real-world data. If you don't believe me, then ask every data professional who transitioned from learning with clean data to cleaning data.

How can we make a simulated environment to be as realistic as possible?

Bring together talented football players and try out different tactics. Let these professionals play out different scenarios and then analyse, perfect, and re-iterate. But here we get into a paradox. We built an AI team to move away from humans in the game, and now we need humans to perfect the AI team. Data comes from humans, and if AI data is used to train itself, we will perfect what is known and never discover anything else.

Let us say that the clubs and the players who train AI agree with all of this. Would you watch the games with the same fever and excitement as you do now? And if your answer is 'no, it won't ever happen', I will ask, ' Why not? ' We are touching every other aspect of humanity, especially privacy, so why not football?